Anda di halaman 1dari 17


discussions, stats, and author profiles for this publication at:

Thermal Analysis—Theory and Applications in


Article in International journal of metalcasting · January 2015

DOI: 10.1007/BF03355598


14 1,444

1 author:

Doru M. Stefanescu
The Ohio State University


Some of the authors of this publication are also working on these related projects:

Nucleation and growth of spheroidal graphite View project

All content following this page was uploaded by Doru M. Stefanescu on 21 January 2015.

The user has requested enhancement of the downloaded file.


Doru M. Stefanescu
The Ohio State University, Columbus, OH, USA
The University of Alabama, Tuscaloosa, AL, USA

Copyright © 2015 American Foundry Society


Monitoring and adjusting the chemical composition and technique evolved then into Differential Thermal Analysis
the nucleation potential of casting alloys melts is critical (DTA) without a reference sample. With this tool in hand
to the production of castings of consistent high quality. the metallurgist ventured to predict not only the chemis-
While other methods such as fractured test samples (e.g., try of the melt but also the nucleation potential (degree of
the wedge test for gray iron) and spectrographic chemi- inoculation), the shrinkage propensity, the dendrite arm
cal analysis deliver useful information, thermal analysis spacing and the grain size, the graphite shape, the solidi-
(TA) (also called cooling curve analysis, CCA) can pro- fication microstructure, and even the room temperature
vide a more complete insight in the dynamic changes oc- microstructure. Currently, different TA techniques are
curring upon melting and during melt treatment of cast- used worldwide in the daily production of all grades of
ing alloys. Initially TA was used for the rapid evaluation cast iron, as well as in monitoring the melting of alumi-
of the carbon equivalent (CE) in cast iron, or the silicon num and steel alloys.
content in Al-Si alloys. Extensive research then extended
the capabilities of TA to the understanding of the solidifi- From this short introduction it should be obvious that the
cation changes induced by compositional variations such story of DTA/CCA is a long and exciting one. This paper will
as the Mn/S ratio, Ce and Mg additions. With the advent try to summarize the fascinating development and extraordi-
of ever faster computers, the first derivative of the cooling nary success of this technique in the casting industry.
curve, which is the cooling rate, was added to the arsenal
of data provided by TA, and a new technique, Computer Keywords: thermal analysis, cooling curve analysis, differ-
Aided Cooling Curve Analysis (CA-CCA), was born. This ential thermal analysis, cast iron, aluminum alloys

Background Cooling curve analysis has evolved over the years becom-
ing a powerful tool for casting process control. The appli-
Thermal Analysis (TA) is the recording and interpreta- cation of CCA to cast iron appears to have started in 1931
tion of the temperature variation in time of a cooling or by Esser and Lautenbusch5 who showed that increasing
heated material. In its simplest form, as applied to met- the superheating of gray iron depresses the eutectic arrest
alcasting, the cooling curve of a metal solidifying in a (Fig. 1). Qualitative observations were also made by Pi-
mold is recorded and analyzed (CCA). Its interpretation wowarski,6 who noticed that higher silicon decreases the
is based on the belief that all the events occurring during undercooling (Fig. 2), by Loper et al.,7 who observed the
solidification leave their mark on the shape of the cooling effect of the Mn/S ratio on the eutectic undercooling (Fig.
curve. It appears that Le Chatelier was the first scientist 3), and by Naro and Wallace8 who revealed the effect of
that recorded temperature as a function of time in heating cerium and sulfur additions (Fig. 4). De Sy and Vidts9
curves in 1887.1 pointed out that cooling curves contain information on
graphite shape.
Differential Thermal Analysis (DTA), which consists of
time, temperature and temperature difference measure- Today, TA can be used to predict alloy composition, grain
ments, appears to have been performed first by Roberts- refining in steel, aluminum, magnesium and other alloys,
Austen2 in 1899, with equipment developments by Kurna- eutectic morphology (e.g., graphite morphology in cast
kov3 and Saladin.4 Differential Thermal Analysis can be irons or degree of modification in Al alloys) and shrinkage
conducted with or without a reference body and using one propensity. Computer analysis of the cooling curve can
or two thermocouples in the same test mold. The refer- provide quantitative information on solidification, such as
ence body is a material, real or virtual (computer-generat- latent heat of solidification, evolution of fraction solid,
ed) that has no phase transformation over the temperature amounts of phases, dendrite coherency and dendrite arm
interval of interest. spacing.

International Journal of Metalcasting/Volume 9, Issue 1, 2015 7

Figure 1. Effect of superheating on the eutectic Figure 2. Effect of silicon on the eutectic temperature of
temperature of gray iron (after Esser and Lautenbusch). gray iron.6

Figure 3. Effect of the Mn/S ratio on gray iron cooling Figure 4. Effect of the Ce and S additions on gray iron
curves.7 cooling curves.8

Direct Thermal Analysis

In its simplest form TA/CCA uses one thermocouple inserted

in the test mold. The design of the test cup has a significant
effect on the results and their interpretation. The shape of the
cooling curve is determined by the balance between the latent
heat liberated during solidification and the heat lost to the sur-
roundings (test cup and atmosphere). It is important that the
design of the cup guarantees consistent sampling conditions.
The cooling curve is affected by the pouring temperature, the
amount of metal poured and the degree of oxidation of the
metal in the cup. A faster cooling because of a smaller test
sample will alter the solidification behavior of the iron, affect-
ing the undercooling. As shown in Fig. 5, as the size of the cup
decreases, undercooling increases. The effect of inoculation is Figure 5. Schematic representation of the effect of cup
more noticeable on the smaller size cups. size and inoculation on the shape of the cooling curve.

8 International Journal of Metalcasting/Volume 9, Issue 1, 2015

Instrumentation and Therminology had the highest undercooling, and using the sand cup,11 CG
exhibited the highest undercooling.
Two types of cups are currently widely used in casting prac-
tice: sand cups and metal cups (see examples in Fig. 6). Sand A combination of number of cups and number of thermo-
cups are cheaper, but a metal cup allows for a more precise couple per cup are used, as follows:
positioning of the thermocouple and more consistent fill-
ing. Figure 7 shows the cooling curves for lamellar graph- • 1 test cup with one thermocouple (TC) (Fig. 6a)
ite (LG), compacted graphite (CG) and spheroidal graphite • 2 test cups, 1 TC per cup
(SG) irons with the same carbon equivalent but poured in • 3 test cups, 1 TC per cup (Fig. 8)
cups of different materials. Note that for the metal cup,10 SG • 1 test cup, 2 TCs (e.g., Fig. 6b)

a) sand cup with disposable thermocouple b) steel cup (SinterCast) with

two thermocouples in the protective tube
Figure 6. Examples of the test cups used for cooling curve analysis.

a) metal cup b) sand cup

Figure 7. Cooling curves for three types of irons obtained with different types of cups with a single thermocouple.

Figure 8. Three cups with one

thermocouple each for evaluation of
Mg content in ductile iron.17

No Te With Te With Te + S

International Journal of Metalcasting/Volume 9, Issue 1, 2015 9

The standard terminology used in CCA is introduced in TLA = 1669 – 124·CEL
Fig. 9 as follows: TL - equilibrium liquidus temperature, Where; CEL = %C + %Si/4 + %P/2 Eqn. 1
TE - equilibrium eutectic temperature, TLA - temperature of
liquidus arrest, TEmin - temperature of eutectic undercooling, A correlation between composition and the liquidus arrest
TEmax - temperature of eutectic recalescence, ∆T - recales- temperature (in °C) was also established by Heine14,15 for hy-
cence, ∆Tmax - maximum undercooling, ∆Tmin - minimum poeutectic gray irons:
TLA = 1569 – 97.3(%C+ 0.25∙%Si) Eqn. 2
Estimation of Chemical Composition Further work16 at BCIRA in England used two cups, a stan-
dard cup and a cup that had a tellurium addition, resulted
Because of the direct correlation between transformation in the ability to determine both C and Si from CCA. The
temperatures and composition, TA has been used exten- tellurium addition had the effect of promoting metastable
sively to build phase diagrams. The first application of (white) solidification which, due to the high growth rate of
CCA in process control was to evaluate the carbon equiva- the white eutectic (ledeburite) all but eliminates the eutectic
lent of cast iron and the silicon content in Al-Si alloys. It is undercooling. The eutectic arrest, TEwhite, becomes flat. The
based on the correlation between the liquidus temperature equations proposed by Heraeus Electro-Nite17 are:
of an alloy (TL) and its chemical composition (%C), as il-
CEL = 14.45 – 0.0089∙TL Eqn. 3
lustrated in Figure 10. However, because solidification is
%C = – 6.51 – 0.0084∙TL + 0.0178∙TEwhite
a non-equilibrium process, the real cooling curve will be
%Si = 78.411 – 4.28087∙Si-adj. – 0.06831∙TEwhite
quite different from the theoretical one. It will exhibit un-
dercooling with respect to the equilibrium temperatures, as
shown on the same figure. Thus, the thermodynamic equi-
librium liquidus temperature will be in most cases higher
than the actual non-equilibrium temperature measured by
the thermocouple.

For the case of cast iron the application of this principle is

based on the concept of carbon equivalent, CE, that includes
the contribution of carbon and of the other important ele-
ments (Si, Mn, P, S, etc.), allowing the multicomponent iron
to be treated like a binary Fe-C alloy. It can be calculated
from equilibrium thermodynamics. However, because of
the discrepancy between the theoretical and actual curve,
Humphreys12 has introduced the Carbon Equivalent Liqui-
dus (CEL) that is different than the thermodynamic eutectic
CE, which is calculated as CE = %C + 0.31∙%Si + 0.33∙%P
– 0.027∙%Mn + 0.4∙%S (see Reference 13 for details). Based
on large number of experiments with standardized test molds
Figure 9. Typical terminology used in TA for casting
(sand cups) the following correlations were established: process control.

Figure 10. Theoretical (equilibrium) and experimental (non-equilibrium)

cooling curves for a hypoeutectic alloy.

10 International Journal of Metalcasting/Volume 9, Issue 1, 2015

Where; Si-adj. is a correction factor, mainly depending on 11a). If the end of solidification is under Tmet the iron is mot-
the phosphorus content of the iron. For SG iron the same tled (mixed gray and white, Fig. 11b). If both the beginning
equation as for LG iron can be used to calculate CEL by and end of solidification are under Tmet, the iron is white (Fig.
substituting TL with (TL + 5). 11c). Note that Tst and Tmet are not straight line, because they
reflect the effect of solute build-up or depletion during so-
A three-cup system was also used to determine the mag- lidification. Carbide promoting elements are rejected in the
nesium content of Mg treated irons. The system includes a liquid, their content increases, and Tmet goes up. The opposite
standard cup, one with tellurium, and one with tellurium and is true for graphite promoting elements.
sulfur (Fig. 8).
Various inoculation additions to the iron greatly affect the eu-
Predicting Eutectic Type and Degree of tectic temperature. As shown in Fig. 12, the addition of 0.002%
Inoculation in Cast Iron Bi lowers the TEmin by 9°F (5°C), and raises the TEmax by 3°F
(l.7°C). A combined addition of 0.002% Bi and a commercial
Another application of single thermocouple TA in cast iron inoculant in the sprue raised the TEmin by 25°F (14°C).18
is for the prediction of the general eutectic structure, i.e.,
gray (stable) or white (metastable). If both the start and end The degree of inoculation of LG iron can be estimated by us-
of eutectic solidification are above the metastable tempera- ing two cups: one before inoculation, and one after inoculation.
ture, Tmet, the iron solidifies gray, i.e., without carbides (Fig. The ratio (∆Tmax) before/(∆Tmax) after steadily increases with
better inoculation (higher number of eutectic grains), as shown
schematically in Fig. 5.

Kanno et al. 19 used three cups with one thermocouple each

to predict the eutectic graphitization ability of LG iron. They
defined the eutectic graphitization ability, EGA, through the
equation (see Fig. 13 for definitions):

Eqn. 4

Here, DTE symbolizes the difference between the stable and

metastable eutectic temperatures. DT1 is the difference be-
a) gray iron
tween the maximum eutectic temperature of the inoculated
iron and the metastable temperature. As shown in Figure 14,
there is a clear correlation between EGA, the type of graphite
and the chill depth. A good correlation was also found be-
tween EGA and the tensile strength, as follows:

TS = (180∙EGA + 170)∙(4.4 - CE) + 160 Eqn. 5

Where; TS is the tensile strength.

b) mottled iron

c) white iron
Figure 11. An interpretation of cooling curves to predict
stable or metastable solidification in cast iron. Figure 12. Effect of Bi inoculation on SG iron.18

International Journal of Metalcasting/Volume 9, Issue 1, 2015 11

Figure 13. Three-thermocouple thermal analysis of gray iron.19

Figure 14. Correlation between the eutectic graphitization ability, type of lamellar graphite
and chilling tendency in lamellar graphite iron.19

Prediction of Graphite Shape ite shape (e.g. Ref. 10, 25, 26). A second derivative of the
T-t curve was then calculated and studied27,28 (see example
One of the early objectives of TA as applied to cast iron in Fig. 18), and then higher order derivatives, as high as the
was prediction of nodularity in ductile iron, with work 5th derivative by Sparkman,29,30 were interpreted with more
being done both in industry (e.g., Ref. 20, 21 and in aca- or less success. While the physical meaning of the first and
demia22). One direction of research focused on the shape second derivatives is clear, that of the higher order deriva-
of the cooling curve (Fig. 15). Another approach was tives is not.
to correlate various temperature data with elements of
the microstructure (Fig. 16). This last method was later Elements of the second derivatives have been used to pre-
extended to the prediction of compacted graphite iron dict microstructure in cast iron and aluminum alloys. For
structure by correlating the eutectic temperature and the example, the following relationship was derived through
recalescence to the graphite morphology, as summa- statistical analysis of the experimental data28 for hypereu-
rized in Fig. 17.10,23 None of these early approaches were tectic (CE = 4.5-4.9) thin wall LG and CG irons with Mg =
deemed accurate enough for process implementation on 0.006-0.031%:
the foundry floor.
Eqn. 6
As early as 1972, a new technique became available for TA
interpretation when Rabus and Polten24 used computer gen-
Where; SN sphericity nodularity defined by:
erated cooling rates from cooling curve data (first derivative
of the temperature-time curve). Derivative calculus permits
the study of the rate of change, and thus of the beginning and
end of transformations. Researchers used combinations of
critical temperature and cooling rate to try to predict graph-

12 International Journal of Metalcasting/Volume 9, Issue 1, 2015

Figure 15. Correlation between the shape of the cooling Figure 16. Correlation between temperatures on the
curves and nodularity.20 cooling curve and microstructure of SG iron.22

a) metal cup10 b) sand cup23

Figure 17. Correlation between temperatures on the cooling curve and microstructure of CC iron.41

Bäckerud et al.38 suggested the use of the maximum of the minum alloys, and the area under it, for the estimation of the
first derivative in the region of primary solidification of alu- efficiency of grain refinement (Fig. 19).

International Journal of Metalcasting/Volume 9, Issue 1, 2015 13

Prediction of Shrinkage Propensity

Some commercial software programs have used the angle of

the cooling rate curve at the end of solidification, GFR 2 in
Fig. 18, as an indicator of the microshrinkage propensity in
cast iron, a large angle suggesting a high shrinkage tenden-
cy.31 This is because the differences in the skin-type solidifica-
tion of LG iron and the mushy-type solidification of SG iron.
Very little latent heat is released at the end of solidification
of LG iron which determines a sharp increase in the cooling
rate. More latent heat is released at the end of solidification of
SG iron, resulting in slower increase in the cooling rate. As
shown in the example in Fig. 20, the lamellar graphite iron
has a higher cooling rate and a narrower angle at the end of so-
lidification when compared with the spheroidal graphite iron.

The use of the first derivative of the T-t curve allows es-
Figure 18. Cooling curve, first derivative and second timation of the time at which various solidification events
derivative.28 are occurring. According to Larrañaga et al.,32 the ratio

a) no grain refinement b) grain refinement

Figure 19. Interpretation of the cooling curve and its first derivative to estimate grain refinement in aluminum alloys.38

14 International Journal of Metalcasting/Volume 9, Issue 1, 2015

k=tEXP/(tExp + tFinal_contr) can use as a predictor for shrinkage TLA = 1594.4 – 102.2(%C+ 0.25∙%Si + 0.5∙%P) Eqn. 7
porosity in cast iron (see Fig. 21 for definitions). A low k is
an indicator of high porosity. The same authors have also For melts superheated at 1510°C (2750°F) the equation was:
proposed an empirical equation for the calculation of nodule
count based on TA data. TLA = 1550 – 92.06(%C+ 0.25∙%Si + 0.5∙%P) Eqn, 8

Prediction of Degree O Oxidation of the Melt These three equations were plotted in Fig. 22. It was confirmed
that the oxygen increases TLA. This effect may be understood
Increased oxygen in the range of 11-101 ppm in an iron melt on the basis of thermodynamics. Oxygen lowers carbon activ-
of composition 3.39-3.46%C and 1.38-1.58%Si has been re- ity, which is equivalent to a decrease in carbon in terms of
ported to raise the liquidus temperature by as much as 10°C solidification behaviour. This implies an increase in the tem-
(18°F).33 Inoculation lowered the TLA. Similar effects were perature. Superheating decreases the liquidus, most probably
noticed for steel where dissolved oxygen raised TLA and de- because it decreases the nucleation potential of the melt.
oxidation with Al depressed TLA.34
Prediction of Mechanical Properties
Heine and Henschel14,35 conducted experiments on a large num-
ber of melts to find the effect of oxygen on the TLA of iron Attempts have also been made to predict mechanical prop-
melts. For melts deoxidized under argon they derived Eqn. 2. erties using TA data by Kano et al. 19 (Eqn. 5) as already
For melts under oxidizing atmosphere the equation was: discussed in this paper. Earlier, Glover et al.36 performed re-
gression analysis on a large number of samples and obtained
the following correlation between tensile strength (TS in psi)
and the liquidus arrest (in °C):

TS = -388,447 + 357∙TLA (R2 = 55) Eqn. 9

The poor correlation coefficient is not surprising, as essen-

tial microstructure features such as graphite shape, amount of
dendrites, dendrite arm spacing, and eutectic grain size have
a rather complex or little influence on the liquidus tempera-
ture. To improve the correlation Glover et al. introduced into
the analysis the dendrite interaction area which represents a
qualitative estimate of the area fraction of dendrites. The new
equation had a higher correlation coefficient, but was still in-
sufficient for reliable prediction for the tensile strength:

TS = -248,504 + 234∙TLA + 93(% Dendrite Interaction)

Figure 20. Comparison of the end of solidification for ductile
(SG) and gray (LG) irons. (R2 = 72) Eqn. 10

Figure 21. Suggested correlation between times and areas from Figure 22. Effect of oxygen and
the cooling and cooling rates curves to microporosity propensity superheating on the liquidus temperature
in cast iron.32 (Heine).

International Journal of Metalcasting/Volume 9, Issue 1, 2015 15

Differential Thermal Analysis (DTA) cannot provide timely data useful in operation. For a more
in depth discussion on classic DTA the reader is referred to
DTA with a Reference Sample
Ref. 37.
Classic DTA is performed with a reference body (Fig. 23a).
Bäckerud et al.38 experimented with a modified DTA in
The sample and the reference body are cooled at a con-
which the reference sample was not a neutral body, but a
trolled, constant speed via the furnace. An example of the
second thermocouple placed in the same cup, close to the
interpretation of the cooling curves is given in Fig. 23b for
wall. Two cooling curves were thus recorded for the same
high speed steel. At the beginning of cooling the tempera-
alloy (Fig. 24a). The temperature difference between the
ture of the sample (Tsam) and of the reference (Tref) follow
wall and center thermocouples that describes the tempera-
parallel paths. When an exothermic reaction, such as a phase
ture gradient across the sample can be used to determine
change occurs, the temperature will increase and a peak will
the occurrence of solidification events such as nucleation at
form on the sample temperature graph. In this case the sud-
wall, dendrite coherency, and formation of various phases
den increase in temperature was produced by the latent heat
such as Mg2Si (Fig. 24b).
released because of the solidification of the δ-phase. Similar
effects are seen for the peritectic reaction and the eutectic
A successful industrial process based on the two-thermocou-
solidification. The areas above the straight line are propor-
ple DTA method is the SinterCast process for production
tional with the latent heat released during the phase trans-
of CG iron.39 It uses a stamped and drawn steel sheet cup
formation. An endothermic peak may occur during cooling
that has a spheroidal containment area (Fig. 6b). The cup is
when a crystalline phase change occurs.
immersed in the molten iron. Two different cooling curves
are obtained from two thermocouples protected by the same
To calculate the latent heat produced during the transforma-
sheath. One of the thermocouples is located in the thermal
tion, we start with the heat flow rate balance of the test - cast-
center of the cup while the second is located close to the
ing/crucible system:
bottom of sample. The walls of the cup are coated with a
reactive coating that consumes active magnesium in order
Eqn. 11

Qf is the solidification latent heat;
t is the time;
v is the sample volume;
ρ is the metal density;
cp is the specific heat of the metal;
T is the average temperature in the sample;
h is the effective heat transfer coefficient;
A is the surface area of the sample;
To is the furnace temperature;
ε is the emissivity of the surface sample;
σ is the Stefan-Boltzmann constant.

It can be shown that, after some manipulations, the time evo- a) equipment for DTA
lution of the latent heat can be calculated as:

Eqn. 12
The total latent heat evolved during solidification, Qf, can
be obtained from the time integration of this equation. The
reaction rate (rate of fraction of solid evolution) can then be
calculated through the numerical integration of:

Eqn. 13

Where; f Si is the fraction solid generated from time zero to

time ti, and Q if is the total heat generated till time i.
b) interpretation of DTA
This method is widely applied in research laboratories, but Figure 23. Schematic representations of the principles
it is not directly used in metalcasting processing because it of DTA.

16 International Journal of Metalcasting/Volume 9, Issue 1, 2015

to simulate the fading of magnesium in the ladle. This al- Where the subscript cc designates the cooling curve and To
lows for the simultaneous measurement of the solidification is the ambient temperature. The thermo-physical quantities
behavior at the start of casting (through the center thermo- are assumed constant. The equation can be rearranged to de-
couple) and also after a predetermined loss of magnesium scribe the cooling rate:
(through the bottom thermocouple). Computer-aided cool-
ing curve analysis allows the prediction of the solidification Eqn. 15
microstructure of the iron and, more importantly, the correc-
tions necessary to achieve the desired CG microstructure. If no phase transformation occurs during cooling Qf = 0 and
the equation becomes:
DTA Without a Reference Sample
Eqn. 16
Integral calculus applied to the temperature-time data allows
evaluation of areas that are directly related to the energy Here the subscript zc denotes the zero-curve, which is the
evolution. DTA lends itself to straight forward integral cal- time evolution of the alloy cooling rate assuming no phase
culus analysis. DTA can also be conducted without a refer- transformation, i.e., the cooling rate of the virtual reference
ence sample. In this case, a test cup with one thermocouple sample. This is the cooling rate of the virtual reference (neu-
is used. The problem is then to calculate the cooling of the tral) body. Assuming that hcc = hzc the time evolution of the
virtual reference sample. There are in principle two major latent heat can be calculated as:
approaches: i) Newtonian analysis and ii) Fourier analysis.
Newtonian analysis requires only one thermocouple and is Eqn. 17
the most widely used. The mathematics is straight forward
(see References 40-47). Fourier analysis is a more accurate and the total heat evolution during solidification is:
treatment of the heat transfer problem, but it requires two
thermocouples and the mathematics is more cumbersome Eqn. 18
(see References 48-50).
The latent heat can be calculated by numerical integration
Newtonian Analysis of this equation:
Eqn. 19
In the Newtonian analysis it is assumed that the thermal gra-
dient across the sample is zero and that heat transfer between
The fraction solid at time i is then calculated with Eqn. 13,
the casting and the mold occurs by convection. The mathe-
or as:
matical background40,41 is similar with that for the DTA with

a reference sample, with some simplifications. Assuming
Eqn. 20
that the heat loss by radiation is negligible during solidifica-
tion, Eqn. 11 becomes:
The problem can now be solved on the Excel spreadsheet.

Eqn. 14

a) cooling curves b) temperature difference between wall and center

Figure 24. DTA with two thermocouples in the same sample. 38

International Journal of Metalcasting/Volume 9, Issue 1, 2015 17

In principle the method consists in generating the first de- sponds to the beginning of the eutectic solidification, while
rivative of the cooling curve with respect to time (the cool- the minimum of the 2nd derivative corresponds to the end of
ing rate), generating a zero-curve, and then subtracting the the eutectic solidification.
area under the zero-curve from the area under the cool-
ing rate. There is no unique accepted method to obtain the To generate the zero-curve (ZC) the heat transfer coefficient
zero-curve. for the cooling without transformation is needed. A possible
approach has been described by Ekpoom and Heine.40 They
The example in Fig. 25 is for a hypoeutectic cast iron. Once calculated a heat transfer coefficient for the liquid stage, hL,
the zero-curve is known, the time evolution of the phase and for the solid stage, hS. In the mushy zone, linear interpo-
transformation is obtained by dividing the area correspond- lation between hL and hS is used. The method is cumbersome.
ing to the phase formation (for example area correspond-
ing to the solidification of primary austenite) to the total Alternatively, the heat transfer coefficient (h) can be gener-
area, the evolution in time of the amount of this phase is ated from the cooling curve or its first derivative. Barlow
obtained. In the example in Fig. 25, if the beginning of eu- and Stefanescu49 proposed the use of a fitted exponential on
tectic solidification is known, the amount of primary aus- the cooling rate before and after solidification, which meant
tenite and the amount of eutectic can be calculated. Then, using one heat transfer coefficient over the entire measuring
the evolution of fraction solid over the solidification time interval. Another procedure was to use three heat transfer
can be plotted. coefficients by fitting exponentials for points on the cool-
ing rate before the beginning of solidification, after the end
There are two problems in the Newtonian analysis: i) estab- of solidification, and another one between the points corre-
lishing the beginning and the end of the austenite and eu- sponding to the beginning and end of solidification.
tectic solidification and ii) calculating the zero-curve. Many
papers suggest the use of the first derivative to establish the In principle, three approaches can be used to compute the
beginning and end of transformation. However, we have zero line (Fig. 27):
demonstrated that this is inaccurate. The beginning and end
of a phase transformation is best obtained from the second 1. ZC1h: logarithmic trend line for points chosen at
derivative of the cooling curve.51 This is because a maxi- the beginning and the end of the cooling rate (CR)
mum of the second derivative indicates a sudden decrease in curve; this method uses only one heat transfer coef-
the cooling rate corresponding to phase solidification, while ficient for the entire cooling curve.
a minimum on the second derivative indicates a sudden in- 2. ZC2h: logarithmic trend line for one point on the
crease in the cooling rate corresponding to the end of any CR corresponding to the beginning of the austenite
solidification (no latent heat production). Figure 26 shows solidification (the maximum of the 2nd derivative)
a cooling curve in the region of the eutectic transformation and points at the end of the CR curve; while only
of an LG iron. The maximum of the 2nd derivative corre- one h is needed for the calculation of the transfor-

Figure 25. Calculation of the amount of phases from the Figure 26. The use of the 2nd derivative to establish the
areas under the cooling rate and the zero curves. beginning and the end of eutectic solidification.

18 International Journal of Metalcasting/Volume 9, Issue 1, 2015

mation region of the ZC, two are needed for the mold can be calculated as , where
whole ZC. T2 and T1 are the temperatures at radii r2 and r1, respectively.
3. ZC3h: logarithmic trend line for points on the CR This introduces the need for two thermocouples. The thermo-
corresponding to the beginning and end of the aus- physical quantities are time and temperature dependent. The
tenite solidification. latent heat and fraction solid evolution are calculated as de-
scribed for the Newtonian analysis. Typical results are shown
Once the ZC is generated, a normalized cooling rate (NCR) in Fig. 29 for an aluminum alloy. The time close to the end
can be plotted. It is the difference between the area under the of solidification, when the cooling curve and the Fourier zero
cooling rate curve and that under the ZC (see Fig. 28). The curve start coinciding, is considered the end of solidification.
normalized cooling rate covers only the solidification part of Note that this occurs earlier than the minimum at the end of
the cooling curve. The area under the normalized CC is pro- the cooling curve.
portional to the latent heat of solidification. The evolution
of the fraction austenite and total fraction solid can then be Mathematically, the Fourier analysis is more accurate than
calculated and plotted. the Newtonian analysis, but its experimental application is
more onerous because the two thermocouples must be po-
Calculated fraction solid obtained from a single-thermocou- sitioned accurately in the measuring cup, an almost impos-
ple Newtonian TA could have significant departure from sible task for sand cups. A more detailed analysis of the Fou-
data obtained through computational modeling of the speci- rier method through inverse heat conduction analysis was
men when the specimen cooled at high Biot number. To in- offered by Diószegi and Hattel.50
crease the precision of TA for cast steel, Lekakh and Rich-
ards46 used a pre-heated thermally insulated ceramic mold.
Fraction solid and dendrite coherency point were evaluated.

Fourier Analysis

Fourier analysis assumes that heat transfer takes place by

conduction only. The following analysis49 closely follows
the method suggested by Fras et al. 48 The Fourier equation
with a heat source term is:

Eqn. 21

with the zero curve given by ZF = ∇2T. To calculate this curve

we must know the temperature field, which for a cylindrical

Figure 28. Normalized cooling rate (N3h) and the

evolution of the solid and primary austenite fractions.51

Figure 27. Three methods of generation of the zero Figure 29. Cooling curve, cooling rate and Fourier zero
curve.51 curve for an aluminum alloy.49

International Journal of Metalcasting/Volume 9, Issue 1, 2015 19

Data Smoothing Techinques Conclusions

An important issue that directly impacts the accuracy and in- Thermal analysis of casting alloys in its direct form or as
terpretation of the DTA is the smoothing of the temperature- differential thermal analysis can provide information about
time data. While there are many smoothing techniques, some the composition of the alloy, the latent heat of solidifica-
of them proprietary, the most common ones are data averag- tion, the evolution of the fraction solid, the amounts and
ing, Gaussian smoothing, and polynomial interpolation.

The simplest technique to implement on the Excel spread-

sheet is data averaging (also known as mean filter) where in
one-dimension the value of a property T over an interval (i
– j) is calculated as . An example of smooth-
ing through averaging using an interval i – j = 9 is provided
in Fig. 30. While no significant differences are noticeable
on the cooling curve, the effect of smoothing on the cooling
rate is marked. Note that some displacement of the maxima
and minima occurs.

Another technique is the Gaussian smoothing. It is simi-

lar to the mean filter but uses the Gaussian distribution
G(x) = , where σ is the standard deviation of the
distribution, to perform the smoothing. The two techniques
are compared in Fig. 31. A Gaussian smoothing add-on for
the Excel spreadsheet was used for the calculation.

The polynomial interpolation is a simple mathematical

procedure that involves fitting of a data set through a poly-
nomial equation of a selected order. An example of fitting
experimental data with a 10th order polynomial is presented
in Fig. 32. A higher order polynomial will provide a better Figure 31. Comparison between Gaussian smoothing
match of the experimental cooling curve. The advantage of (dotted lines) and averaging smoothing (full lines). The
this method is that the data is completely smooth, facilitating data for averaging have been displaced to the right by
the detection of the critical points on the cooling curve. 20s on purpose, to facilitate observation of differences.

Figure 30. Comparison between the raw data(dotted Figure 32. Comparison between the raw data(dotted
lines) and data smoothed by averaging over 9 points lines) and data smoothed by 10th order polynomial
(full lines). interpolation (full lines).

20 International Journal of Metalcasting/Volume 9, Issue 1, 2015

types of phases that solidify, and even dendrite coherency. 19. T. Kanno, I. Kang, Y. Fukuda, M. Morinaka and H.
There are also many other uses for TA, such as, determining Nakae, AFS Trans. (2006) paper 06-083.
dendrite arm spacing in aluminum alloys (e.g., References 20. E.F. Ryntz, J.F. Janowak, A.W. Hochstein and C.A.
52,53) and steel,46 degree of modification and grain refining Wargel, AFS Trans. 79 (1971) 161-164.
in aluminum alloys (e.g., References 54-57 in addition to 21. R. Monroe and C.E. Bates, AFS Trans. 90 (1982) 307-311.
work already cited), graphite morphology and the degree of 22. M.D. Chaudhari, R.W. Heine and C.R. Loper, AFS
nodularity in cast irons (e.g., References 58,59 in addition to Trans. 79 (1971) 399.
work already cited), and even determining the most effective 23. D.M. Stefanescu, C.R. Loper, R.C. Voigt and I.G.
heat treatment cycle for the production of austempered duc- Chen, AFS Trans. 90 (1982) 333.
tile iron.60 The use of TA for process control in metalcasting 24. D. Rabus and S. Polten, Giesserei Rundshau, no. 9
is very extensive and dynamic ongoing, with improved in (1972) 1-8.
equipment and data analysis. 25. P. Strizik, Giesserei 61 (1974) 615-618.
26. D.M. Stefanescu, L. Dinescu, S. Craciun and M.
The metalcaster should be aware that the vast majority of Popescu, in Proc. 46th Int. Foundry Cong., Madrid,
the information provided by the commercial software for TA Spain, CIATF, (October 1979).
and DTA is based on empirical equations obtained through 27. I.G. Chen, D.M. Stefanescu, AFS Transactions, 92
statistical analysis, as shown through examples in this paper. (1984) 947-964.
Therefore, the input of the plant metallurgist is a necessary 28. S. Charoenvilaisiri and D. M. Stefanescu, Proceedings
prerequisite for confidence building in the tools used for TA. of the 7th Asian Foundry Congress, Taipei, Taiwan
(2001) 91.
REFERENCES 29. D.A. Sparkman, “Thermal Analysis Metrics by
1. H. Le Chatelier, Z. Phys. Chem. 1 (1887) 396. feb2010htopic.pdf (last accessed 10-30-14).
2. W.C. Roberts-Austen, Metallographist, 2 (1899) 186. 30. D.A. Sparkman, Trans. AFS 119 (2011).
3. N.S. Kurnakov, Z. Anorg. Chemie, 42 (1904) 184. 31. P.E. Persson, A. Udroiu, P. Vomacka, Wang Xiaojing
4. E. Saladin, Iron and Steel Metallurgy Metallography, 7 and T. Sjögren, Proceedings of the 69th World Foundry
(1904) 237. Congress (2010).
5. E. Piwowarski, Hochwertiges Gusseisen, Springer 32. P. Larrañaga, J.M. Gutierez, A. Loizaga, J. Sertucha,
Verlag, Berlin (1961). and R. Suárez, Trans. AFS 116 (2008) paper 08-008.
6. E. Piwowarski, Giesserei 25 (1938) 523. 33. N. Kayama, K. Murai, H. Nakae, Report Casting Res.
7. C.R. Loper, R.W. Heine and A. Shah, AFS Trans. 75 Lab., Waseda Univ., Tokyo no.21 (1970).
(1967) 541. 34. G. Mahn, B.C. Mahanty, Stahl und Eisen, 89(23)
8. R.L. Naro and J.F. Wallace, AFS Trans. 78 (1970) (1969).
229-238. 35. C. Henschel, R.W. Heine, AFS Cast Metals Res. J.
9. A. De Sy, J. Vidts, Traité de metallurgie structurale (Sept. 1971).
theorique et appliquée, Ed. Dunod, Paris (1962) 464p. 36. D. Glover, C.E. Bates and R. Monroe, Trans. AFS, 90
10. L. Bäckerud, K. Nilsson and H. Steen, in The (1982) 745-757.
Metallurgy of Cast Iron, B. Lux, I. Minkoff and F. 37. H. Fredriksson and Ulla Åkerlind, Solidification and
Mollard, Eds., Georgi Publishing, Switzerland (1975) Crystallization Processing in Metals and Alloys, Wiley
625. (2012).
11. D.M. Stefanescu, in The Physical Metallurgy of Cast 38. L. Bäckerud, G.C. Chai and J. Tamminen,
Iron, H. Fredriksson and M. Hillert eds.,Elsevier Solidification Characteristics of Aluminum Alloys,
(1985) 151. AFS/Skanaluminium, Stockholm (1990) 256p.
12. J.G. Humphreys, BCIRA J. 9 (1961) 609-621. 39. S. Dawson, P. Popelar, in 2013 Keith Millis
13. D.M. Stefanescu and S. Katz, “Thermodynamic Symposium on Ductile Iron (2013) 59.
Properties of Iron-Base Alloys,” ASM Handbook vol. 40. L. Ekpoom and R. W. Heine, Trans. AFS, 89 (1981)
15 – Casting (2008) 41-55. 27.
14. R.W. Heine, AFS Cast Metals Res. J. (June 1971) 49- 41. I.G. Chen and D. M. Stefanescu, Trans. AFS 92 (1984)
54. 947.
15. R.W. Heine, AFS Trans. 82 (1974) 462-470. 42. S. L. Bäckerud and G. K. Sigworth, Trans. AFS 97
16. W. Donald and A. Moore, BCIRA Report no. 1128 (1989) 459.
(1973). 43. K.G. Upadhya, D. M. Stefanescu, K. Lieu and D.P.
17. W. Van der Perre: “Thermal analysis, principles and Yeger, Trans. AFS 97 (1989) 61.
applications,” Heraeus Electro-Nite website 44. W.T. Kierkus and J. H. Sokolowski, Trans. AFS 107 (last accessed 10-30- (1999) 161.
14). 45. R. Mackay, J. H. Sokolowski, R. Hasenbusch, and
18. G.R. Strong, AFS Trans. 91 (1983) 151-156. W.J. Evans, Trans. AFS, paper 03-96 (2003).

International Journal of Metalcasting/Volume 9, Issue 1, 2015 21

46. S. N. Lekakh and V. L. Richards, Trans. AFS 119 53. S. Gowri, AFS Trans. 102 (1994) 503-508.
(2011) paper 11-042. 54. D. Apelian, G.K. Sigworth and K.R. Whaler, AFS
47. A. Dioszegi and I. Svensson Mat. Sc. Eng. A 413-414 Trans. 92 (1984) 297-307.
(2005) 474. 55. J. Charbonnier, AFS Trans. 92 (1984) 907-922.
48. E. Fras, W. Kapturkiewicz, A. Burbielko and H.F. 56. D. Apelian and J.J.A. Cheng, AFS Trans. 94 (1986)
Lopez, Trans. AFS 101 (1993) 505. 797-808.
49. J.O. Barlow and D.M. Stefanescu, Trans. AFS 105 57. N. Tenekedjiev and J.E. Gruzleski, AFS Trans. 99
(1997) 349. (1991) 1-6.
50. A. Dioszegi and J. Hattel, Int. J. Cast Metals Research 58. D. Sparkman, AFS Trans. 102 (1994) 229-233.
17 (2004) 311. 59. P. Zhu and R.W. Smith, AFS Trans. 103 (1995) 601-
51. G. Alonso, P. Larrañaga, J. Sertucha, R. Suárez, D.M. 609.
Stefanescu, Trans. AFS 120 (2012) 329-35. 60. C.H. Chang and T.S. Shih, AFS Trans. 102 (1994)
52. L. Ananthanarayanan, F.H. Samuel and J.E. Gruzleski, 357-365.
AFS Trans. 100 (1992) 383-391.

22 International Journal of Metalcasting/Volume 9, Issue 1, 2015

View publication stats